Apache Cassandra Apache Cassandra World Party | July 2022, 20 Jul 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: What is Netflix waiting for in Apache Cassandra 4.x? | Apache Cassandra World Party 2022

Description

Presentation from the second Cassandra World Party on July 20, 2022. Learn more about Apache Cassandra at https://cassandra.apache.org/

Vinay Chella of Netflix presents “What is Netflix waiting for in Apache Cassandra 4.x?”
This talk covers the newly built Cassandra-based systems at Netflix, its plan for adopting Cassandra 4.x, and the features we are most excited about and looking forward to in Cassandra 4.x.

About the Presenter
Vinay is an Engineering Leader at Netflix and has been involved with Apache Cassandra for 10 years including being a committer for the project.

Apache Cassandra is an open source NoSQL distributed database trusted by thousands of companies for scalability and high availability without compromising performance. Linear scalability and proven fault tolerance on commodity hardware or cloud infrastructure make it the perfect platform for mission-critical data. Get started: https://cassandra.apache.org/quickstart/

The Apache Cassandra World Party 2022 was a virtual, one-day celebration on Wednesday, July 20, 2022 held in anticipation of the launch of 4.1

Join the Cassandra community: https://cassandra.apache.org/community/

#ApacheCassandra #database #NoSQL

A

Well, um thank you, everyone and thank you, for uh you know arranging this uh amazing apache cassandra world, partly I'm so excited to be here. My name is uh I've been practicing uh cassandra for the last 10 years. Since you know 0.7 version, I'm also apache center committer and engineering leader of netflix.

A

There are a lot of exciting features coming in forex, but this talk covers some of the features that we at netflix are particularly excited about and looking forward to uh in this talk, I present an overview of just some of the many implements that are shipping in forex um next slide. Please a brief history of cassandra usage at netflix we've been using cassandra for over a decade there. um It has helped us build a seamless globally, always on experience for our users, and it has also presented itself as a de facto database.

A

For you know, 99 of uh streaming persistent data um next slide um a typical use cases on uh cassandra. Instead of going through. You know a long list of use cases I thought of why don't I share about very uncommon or typical use cases that you ever heard of being built on top of cassandra and if you have already doing that, that is great.

A

Some of them are, you know, starting with storing microservices data, which is like petabyte scale and lots of big payloads, um storing device logging data, which is very high traffic use case uh billing and payments use cases which is very complex in terms of functionality and complaints, um and also like pi data great now that we have seen where netflix uses cassandra. Let's take a quick look at what are the you know? What are we at netflix most uh excited about the next slide?

A

Please um we try to categorize that into three broader buckets uh next slide. Please.

A

We try to categorize that into three broader buckets, uh reliability, correctness and new features. Next slide, please to start with in terms of reliability, one of the uh significant improvement um is in the internal networking. Like so far, clients have were already fully async right, uh especially with the sql but however, now server to server communication is also sync as well. You know with the introduction of uh non-non-blocking io messaging service, we don't have any more thread per code um threat by connection architecture.

A

It is fully async which is close to 20 faster and also enable cassandra to use uh some of the most critical os networking features.

A

Now we can also safely restart cassandra servers without impacting high throughput, bigger sized cassandra clusters. 99 latencies. Numerous improvements have been made, such as like not accepting requests without establishing fully functional connection to its peers and not restarted, not sending the request to the you know, dead notes for a while, which is basically impacts. All your tail latencies next slide. Please, um cassandra. 4X um also has the ability to deny list the bad partitions when, um when it is either being ddosed or causing nice noisy neighbor problems on your multi-tenant clusters?

A

Next slide, please so various other improvements have also been made uh to improve the speculative reflex when there are slow nodes, especially reducing the number of you know, traditional number of v nodes next slide. Please um number of me notes to improve the availability with more profound research. You know and analysis on, especially introducing the number of v nodes um you know. Also, there is a incremental repair work that has uh uh that has that has made tremendous progress in photo, which is especially in the correctness aspect.

A

I think we are still on the previous slide.

A

All right, yeah, so incremental repair in forex, actually works. Now, with a lot of improvements made informative repair. Now we can run them. You know safely on, like petabyte-sized uh data sets much faster and safer.

A

This also opens up opportunities to you know for the operators to preview the data inconsistencies before they actually get repaired uh and remember the cassandra is hopefully consistent without repair, so which is what you don't expect from a database, so in 4a in 4x, um you can run this request much faster on on a bigger data sets uh next on the availability aspect. uh We're talking about you know. Internet networking also helped us in increasing the availability point of view in cassandra forex, especially with the introduction of uh zero copy transfer.

A

uh We can now stream replace or you know, start a bring a new node into the cassandra ring much faster like when I talk about you, know speed. We are literally maxing your instances, network capacity, uh which is improving your mean time to recover like when there is an incident and stuff like that.

A

By like fivefold, um a bunch of new features are also being shipping being shipped in for it forex, uh but I mentioned some of them here like starting with audit logging, which is basically a long awaited and most demanded enterprise feature uh from a pci and compliance partner like with this feature on forex now, you can safely check that box on your db compliance list, and you can use cassandra for most secure to store your most secure data and in your enterprise in your enterprises.

A

um Cassandra forex also makes it easy to operate your you know shared cassandra clusters by providing more insights um through virtual tables and stuff like that. Well, you know what can you walk away with? Well, you know patrick reminded me dude. This is not a in-person meeting. This is the virtual.

A

Well, what can you log off with today, uh if you're not using cassandra yet start with forex today, if you're already using cassandra next slide start upgrading to forex, which is basically uh clustered by tested by uh you, know the biggest and most prominent deployments of the cassandra world, so you are in the good and safe hands.

A

um Thank you. uh Thank you. Patrick, thank you katrina. Thank you for arranging this. This is great thanks. Vinay, it was great to hear from you today was able to answer why we are waiting.

A

uh What are we excited about? Well, I was busy retweeting from patrick I'm sorry, oh, my goodness go check your twitter.

A

Next time I better send a bio yeah next time bring a bio, that's all yeah. I will make one up for you, but it was pretty good.

A

It was great. Thank you. Thank you.